San Jose, California-based Synaptics Inc, a pioneer in the field of neural network chips, has announced the incorporation of its electronic ‘eye’, the l-1000 Neural Eye Chip, into VeriFone Inc’s new Gemstone Onyx cheque identification system. The chip enables the Onyx device actually to see and recognise the identification numbers on cheques. This gives the system a number of advantages over its magnetic counterparts: it does not require cheques to be machine-fed, at a constant speed and angle; it is not stumped by differing ink densities or imperfections in print alignment, and it doesn’t blink an eye at crumpled or folded cheques. It is claimed to cost up to 100 times less than other current systems too. The l-1000 chip is modelled on the synapses of the human brain. It combines an analogue optical sensor, a neural network that recognises and learns patterns, and a signal processor. It scans characters and sends signals to its image classifier, an on-chip network of 20,000 analogue processors, which finds and interprets patterns in the analogue signals. Recognised patterns are relayed to the chip’s processing system, which converts them to a digital format. The chip can read up to 1,000 characters per second and can perform up to 1,000m operations per second. Synaptics plan to manufacture it in volume and sell it on the open market envisaging, as technology progresses, that l-1000-type processors will be used for a variety of applications from the detection of counterfeit money to handwriting recognition (Synaptics is researching a system for deciphering handwritten enevelopes for the US Postal Service, CI No 1,954). A telephone translation device capable of converting one language to another and a neural interface between humans and computers are also suggestions for the long-term future. When you combine the two technologies, logical and intuitive, you will be able to make machines that you cannot conceive of today, says Synaptics’ co-founder Frederico Faggin.
